Introduction

Ischemic leukoaraiosis (IL) is a cerebral small vessel disease defined as a combination of diffuse white matter (WM) lesions (leukoaraiosis) and a history of clinical lacunar stroke [1]. The WM changes appear as hyperintense lesions in T2-weighted (T2w) MR images. Furthermore, there is evidence of histopathological rarefaction of myelin and axons due to alterations of the small penetrating arteries and arterioles, of the veins, and microcirculatory disturbances as well as activation of the coagulation-fibrinolysis pathway [2–6]. The alterations of small arteries and veins result in increased vessel length and narrowed lumina, yielding a decrease of perfusion pressure and cerebral blood flow (CBF) [1, 3, 5, 7–14]. Although a subsequent increase of the oxygen extraction fraction (OEF) partly compensates for the CBF reduction [15], there is a higher risk of hypoxia.

Cerebral small vessel disease clinically manifests as lacunar stroke with focal neurological deficits, being the most common cause of vascular dementia [7, 11, 12, 14, 16–25]. However, the correlation between the lesion load, as assessed from conventional T2w MR images, and clinical parameters remains weak [1, 11, 12, 16, 26–28]. Moreover, increased signal intensity in standard MRI may reflect a spectrum of pathological abnormalities ranging from very mild tissue changes to complete demyelination and axonal loss [12].

In contrast to conventional MRI, quantitative MRI techniques are more sensitive to microstructural and metabolic changes in brain tissue and might therefore differentiate between normal age-related white matter changes and potentially harmful changes due to small vessel disease [12, 14, 16, 27, 29–33].

The transverse relaxation time of blood water strongly depends on blood oxygenation, as spin dephasing induced by magnetic field distortions around erythrocytes increases in the presence of paramagnetic deoxygenated hemoglobin (deoxyHb). The resulting magnetic susceptibility differences between blood vessels and surrounding tissue reduce the effective transverse relaxation time T2*, and thus the signal in T2*-weighted images. However, T2* is also affected by changes of the transverse relaxation time T2, which increases in the cerebral cortex and thalami with age and in many cerebral pathologies due to edema or gliosis. To account for this effect, the relaxation time T2′ may be calculated from 1/T2′ = 1/T2* − 1/T2. T2′ provides mainly information about local concentrations of deoxyHb [34–36].

High-resolution, motion-corrected quantitative mapping of T2, T2*, and T2′ depicts microstructural changes and the oxygenation status in healthy individuals of younger and older ages [15, 37] and in patients with high-grade stenosis of intracranial arteries [38]. Likewise, combined mapping of the different transverse relaxation times T2, T2* and T2′ should also help to assess the grade of (micro-)structural damage (T2) and hypoxia (T2′), even in the normal appearing WM and gray matter (GM) of patients with IL. Here, respective mapping techniques were combined with pulsed arterial spin labeling (PASL)-based perfusion mapping of cortical CBF and quantification of cortical GM and WM atrophy to gather a more complete picture of IL.

Methods

Subjects

Eighteen patients with IL (age range 46–89 years; mean 71 years) were recruited via the local Department of Neurology. Inclusion criteria were as follows: (1) extensive WM lesions on conventional CT or MRI (Fazekas grade II or III) and (2) a recent (<2 weeks) clinical history of lacunar stroke. Furthermore, 18 age-matched healthy controls (age range 46–88 years; mean 67 years; comparison with patients, p = 0.215) were included in the study. Controls were required to show no or only mild WM lesions (Fazekas grade 0 or I), and to have no clinical history of stroke [21]. Standardized questionnaires and medical records were used to exclude other neurologic or psychiatric diseases including manifested dementia, diabetes, or untreated hypertension in both groups.

The study protocol was approved by the Ethics Committee, Faculty of Medicine, Goethe University Frankfurt, and informed consent was obtained from each subject.

MR examination

Scans were performed on a 3-T MR scanner (Magnetom Trio; Siemens Medical Solutions, Erlangen, Germany), equipped with an eight-channel phased-array head coil for signal reception and a body coil for radio frequency (RF) transmission.

As described in detail elsewhere [15, 37], for quantitative T2* mapping, a multiple gradient echo sequence with eight echoes per excitation acquired exclusively under positive readout gradient polarity [39] was employed with the following imaging parameters: TE ranging from 10 to 52 ms with a constant increment of 6 ms, TR = 3000 ms, 50 axial slices with 2 mm slice thickness, no interslice gap, bandwidth = 300 Hz/pixel, excitation angle 30°, matrix size 160 × 128 (readout × phase encoding), field of view (FoV) 200 × 160 mm2, isotropic spatial in-plane resolution of 1.25 × 1.25 mm2, measurement time 6 min 24 s. To avoid erroneous T2* values due to distortions of the static magnetic field B0 [40–42], a compensation method was applied, based on obtaining B0 maps directly from the complex image data and correcting for TE-dependent image intensity reductions induced by spin dephasing [43]. The acquisition was repeated with 50 % spatial resolution in phase encoding direction (duration 3 min 30 s), and a special combination of both data sets in k-space was performed for the suppression of motion-induced artifacts [44].

Quantitative T2 mapping was based on the acquisition of a series of T2-weighted images with different values of TE (17, 86, 103, 120, 188 ms) but otherwise identical imaging parameters, using a turbo spin echo (TSE) sequence with TR = 10 s, bandwidth = 100 Hz/pixel, turbo factor = 11. The relevant geometrical parameters (slice coverage and in-plane resolution) were identical to the T2* mapping experiment. The duration of each single T2-weighted data set was 2 min 22 s, so the total duration of T2 mapping was 11 min 50 s. Before T2 fitting, the TSE images were coregistered to the first T2*-weighted image using the linear registration algorithm implemented in FSL (FLIRT, FMRIB’s Software Library-FSL, http://www.fmrib.ox.ac.uk/fsl). Data fitting was performed with custom-built programs written in MATLAB (MATLAB, The Mathworks Inc., Natick/MA, USA), using mono-exponential fitting for obtaining maps of T2* and T2. Maps of T2′ were calculated as 1/T2′ = 1/T2* − 1/T2.

Anatomical imaging was based on a T1-weighted MPRAGE sequence [45] with the following imaging parameters: TR/TE/TI = 2250 ms/2.6 ms/900 ms, FoV 256 × 224 × 144 mm3, whole brain coverage, isotropic spatial resolution 1 mm, 22 % oversampling in slice encoding (left-right) direction to avoid aliasing, bandwidth = 200 Hz/pixel, excitation angle 9°, duration 7:23 min. This data set was used for segmentation of GM and WM. The fitted maps (T2, T2*, and T2′) were linearly coregistered to the T1-weighted anatomical data set using FLIRT. A 3D fluid attenuation inversion recovery (FLAIR) scan was also acquired to better differentiate lesions from normal appearing WM, GM, or cerebrospinal fluid (CSF), using the parameters TR/TE/TI = 6000 ms/353 ms/2200 ms, FoV 250 × 215 × 160 mm3, whole brain coverage, isotropic spatial resolution 1 mm, bandwidth = 930 Hz/pixel, duration 8:08 min.

For quantitative measurement of brain perfusion, PASL-based CBF maps were acquired in 15 patients and 15 age-matched controls, sampling 100 pairs of control and tag images. Labeling was based on Q2TIPS (quantitative imaging of perfusion using a single subtraction, second version with interleaved thin-slice TI periodic saturation) as described by Luh et al. [46]. The proximal inversion with a control for off resonance effects (PICORE) labeling scheme was applied, labeling arterial spins inside a thick inversion slab proximal to the imaging slices. The scanning parameters were as follows: field of view (FOV) = 240 × 240 mm2, matrix size = 64 × 64, isotropic in-plane resolution of 3.75 × 3.75 mm2, five axial slices with 6 mm thickness placed at the centrum semiovale (Figs. 1 and 2), interslice gap of 2 mm, TR = 2500 ms, TE = 21 ms, TI1/TI1stop/TI2 = 700 ms/1375 ms/1400 ms, bandwidth = 2694 Hz/pixel. The gap between the tag inversion slab and the first of the imaging slices was 10 mm. The duration was 8:20 min.

Fig. 1
figure 1

MR images from a patient. Upper row shows T2 (A), T2* (B), and T2′ (C) maps. Regions of interest used in the study (D): lesions (red) are manually drawn on the FLAIR images (E, arrows), while gray matter (gray) and white matter (white) are automatically segmented based on T1-weighted (T1-w) images (F). An additional region of interest (blue) was manually drawn on the CBF maps (PASL) (I) on five slices at the level of the centrum semiovale as demonstrated on sagittal (G) and coronal (H) views (overlapped on T1w image). Note the inhomogeneous gray values distribution of the quantitative maps that makes visual assessment difficult and demonstrates the necessity of quantitative evaluation. This slice is compatible with patient’s group results of T2 and T2* values but is an outlier for T2′ values in the lesions and the GM (higher than group results)

Full size image
Fig. 2
figure 2

MR images from a control subject. Panel’s description as in Fig. 1. This slice is compatible with controls’ group results for T2* and T2′ values but is an outlier for T2 values in the lesions and the GM (higher than group results)

Full size image

Hyperintense lesions were manually delineated as regions of interest (ROIs) on the FLAIR data set and were further coregistered to the T1-weighted anatomical data. Due to the limited field of view (five slices) in the case of PASL, coregistration to the reference anatomical images was not possible; thus, differentiation of CBF in cortex and subcortical structures was not possible, and PASL-GM ROIs had to be defined directly on the calculated CBF maps (Figs. 1 and 2).

Cortical GM and whole WM ROIs were defined following automatic segmentation with SPM 8 (http://www.fil.ion.ucl.ac.uk/spm) of the T1-weighted data set by classifying each voxel as either GM, WM, or CSF, depending on the highest probability value in the respective tissue maps, and excluding the lesions ROIs. Manually (lesions) and automatically (WM and GM) derived ROIs were used as masks to extract T2, T2*, and T2′ values from the corresponding maps, while the PASL-GM ROI was used to extract PASL-based CBF values (Figs. 1 and 2).

Relative volume fractions of GM, WM, and lesions were calculated, dividing the respective volume by the total intracranial volume (the latter being calculated as the combined volume of GM, WM, CSF, and lesions). It should be noted that GM, WM, or CSF compartments were defined as mutually exclusive regions, each voxel being assigned to one tissue class with exclusion of the manually defined lesion ROIs as explained above.

All parameters derived as described above were compared between patients and healthy controls using t testing. Statistical comparison was performed by means of the software R Statistics 2.15. Values of p < 0.05 were accepted as statistically significant.

Results

Within the lesions and the unaffected WM, the T2 relaxation time was significantly increased in patients as compared to controls (p < 0.01 and p = 0.045, respectively). In GM, there were no T2 differences.

The T2* relaxation time was significantly increased in the patients as compared to controls in the lesions (p = 0.003). However, there were no T2* differences in WM or GM.

T2′ was significantly decreased in normal appearing WM of patients (p = 0.027). In contrast, there were no T2′ alterations in patients’ GM as compared to controls (Table 1).

Table 1 Values of relaxation times T2, T2*, and T2′ [ms], comparison between patients with IL and controls in three regions of interest: lesions, white matter (WM), and gray matter (GM)
Full size table

No changes could be found in the PASL-based CBF in GM between patients and controls (53.06 ± 21.85 ml/100 g/min and 48.43 ± 12.87 ml/100 g/min, respectively, p = 0.49). However, it is worth noting that the statistical scattering was smaller in the control group.

As expected from inclusion criteria, the lesion fraction was significantly larger in patients than that in controls (p < 0.01). Patients showed a significant smaller fraction of normal appearing WM and GM (lesions excluded) than controls (p < 0.01) (Table 2).

Table 2 Fractions of lesion, white matter (WM), and gray matter (GM) relative to total intracranial volume, comparison between IL patients and controls
Full size table

Discussion

This study employed quantitative T2, T2*, and T2′ mapping, PASL-based perfusion mapping, and WM-GM segmentation to assess (micro-)structural damage and hypoxia in lesions and normal appearing WM and GM in patients with IL compared with age-matched healthy controls.

Results (Table 1) show that in patients with IL, T2 is significantly increased both in lesions and in the normal appearing WM. In contrast, no T2 differences were found in GM.

Hyperintense lesions on standard T2-weighted MRI are rather unspecific and can reflect a spectrum of abnormalities, ranging from very mild tissue changes to complex demyelination and axonal loss in the presence of small vessel disease [12]. In patients with IL, WM lesions are caused by lacunar or incomplete infarction with a damage of myelin, axons, and astroglia [3–6, 29, 47]. In contrast, WM lesions appearing during normal aging [15, 33]—as in our control group—appear to be of non-vascular origin without tissue damage but with reduced myelination, atrophy of the neuropil, and enlarged perivascular spaces [6, 27]. While this difference is not sufficiently assessable via conventional MR imaging, quantitative T2 mapping might help to distinguish the nature of these lesions.

The observed T2 increase in the normal appearing WM (after lesion exclusion) of IL patients as compared to age-matched controls may be caused by diffuse WM damage with demyelination and microinfarcts in IL patients [4, 28]. Previous studies using quantitative T2 mapping revealed a T2 increase with age, even in healthy volunteers; however, WM lesions were not excluded from WM analysis in these studies [15, 33]. This global WM damage—even in the normal appearing WM—could already be demonstrated using different quantitative MR techniques, including magnetization transfer (MTR) mapping [27, 32] and diffusion tensor imaging (DTI) [12, 16, 29–31]. In these studies, changes of the normal appearing WM of patients with vascular changes corresponded to damaged ultrastructure even outside of the visible lesions. Further, the observed changes correlated with clinical presentation. Assuming that T2 prolongation also reflects microstructural damage (see also below), quantitative T2 mapping might therefore be helpful to assess prognosis and to monitor therapeutic effects in patients with IL.

In the GM of patients, one might have expected an increase of T2 due to the parenchymal damage caused by cortical microinfarcts. However, no T2 alterations were found (Table 1), probably due to a good collateral supply saving the GM from substantial damage [4, 19, 28].

T2* was significantly increased in the lesions of IL patients as compared to controls, while no T2* changes were observed in the normal appearing WM or in the GM (Table 1). T2* is influenced by paramagnetic tissue properties (T2′) which primarily depend on heme-bound and ferritin-bound iron. However, T2* also depends on diamagnetic effects (T2), which are more marked in the WM and in patients with IL. To assess the paramagnetic tissue properties without the influence of T2 effects, T2′ should be used (see below) [15, 37, 48]. Since T2′ was unaltered in the lesions of patients as compared to controls, the significant increase of T2* in the lesions of patients is likely an effect of the locally increased T2.

In the GM, T2′ was similar in patients and in controls (Table 1). Again, the good collateral supply may save the GM from hypoxia in patients with IL [4, 19, 28].

However, in the unaffected WM, T2′ was significantly decreased in patients (Table 1). As T2′ decreases with increasing concentrations of paramagnetic deoxyHb, T2′ serves as a marker for hypoxia [15, 37, 38]. The decreased T2′ in the WM of patients with IL therefore seems to reflect the hypoxic state in their normal appearing WM. Misery perfusion might be one important factor contributing to hypoxia in IL patients. Due to pathologic changes of the small arteries and veins and the decreased vascular density in the WM of IL patients, the reduced hemodynamic response is not sufficient to meet the metabolic demands in the case of even small decreases in cerebral perfusion pressure. As a consequence, OEF increases [3, 10, 11]. An increased OEF along with reduced CBF, however, is associated with an increased risk of incomplete infarction or stroke [3, 49, 50]. As described by Schmidt et al. [6], factors induced by hypoxia can be found in deep and subcortical WM lesions in patients with cerebral microangiopathy. Previous studies using positron emission tomography (PET) with different tracers showed partly contradictory results. While OEF was unaltered in patients with vascular dementia or non-demented patients with leukoaraiosis in some PET studies [17, 51], other authors reported OEF increases [10, 11]. Meguro [52] showed that OEF was higher in patients with severe periventricular hyperintensities (PVH) compared with those with mild or no PVH, although lacking statistical significance.

Therefore, the decrease of T2′ in normal appearing WM might be interpreted as a marker of hypoxia and might help to estimate the risk of new lesions in patients with IL. Furthermore, Yao et al. [11] stated that cognitive function in patients with vascular dementia does not directly correlate with the severity of WM lesions but rather with a decreased oxygen metabolism detected by PET. Therefore, T2′ mapping might also be useful to assess the risk of cognitive impairment in patients with IL.

Further, the observation of both increased T2 and decreased T2′ in the normal appearing WM might support the hypothesis of severe structural damage in patients with IL. Deoxygenation has a bearing on T2, and hypoxia leads to a T2 decrease, especially in the blood [53, 54]. However, tissue T2 is also influenced by certain pathologies such as the occurrence of edema or gliosis which yield a T2 increase. In contrast, T2′ is primarily affected by paramagnetic tissue properties and is therefore a more sensitive marker of hypoxia. If, as in this study presented here, both a decrease in tissue T2′ (as a marker of hypoxia) and an increase in tissue T2 are observed, the latter effect is likely to be caused by pathological tissue alterations that level out the decreasing effect of hypoxia on T2. Thus, the combination of increased T2 and decreased T2′ should indicate even more severe tissue damage.

We found no difference in cortical CBF between patients and controls on the basis of PASL-based CBF maps. Other studies found a cortical CBF decrease in IL patients with different imaging modalities including PET and MRASL [52, 55]. This discrepancy in results may be due to the use of different mapping techniques. It has been shown that PASL may lead to an overestimation of CBF in patients with small vessel disease like IL, which might be caused by stagnation of labeled blood in small vessel ectasia [56]. Furthermore, in the case of decreased perfusion pressure, cerebral autoregulation might induce a cerebral blood volume (CBV) increase which also might contribute to a PASL-based overestimation of CBF in IL. In our IL patients, PASL-based CBF values showed a wider scattering than for controls, probably due to partial volume effects resulting from the significant decrease of cortical volume (see below) [56, 57]. Unfortunately, CBF in the WM cannot be reliably determined due to the low signal to noise ratio even at 3 T, which prevents detection of the almost fourfold lower perfusion in white matter relative to gray matter [37, 58]. However, it has been shown that there is a CBF decrease in WM of IL patients, and we speculate that this may also be true for our study [1, 3, 5, 7–14].

In fact, we found significantly lower GM and WM fractions in IL patients than those in controls (Table 2). This is in line with the results of previous studies that reported cortical atrophy in patients with leukoaraiosis and subcortical vascular disease [13, 59]. The mechanism of cortical atrophy in patients with IL is not yet fully elucidated and might be of primary or secondary nature [13, 18, 60]. Synaptic loss, apoptosis, necrosis, shrinkage, and decreased CBF [14, 18] as well as direct injury from microscopic cortical infarcts [60] might contribute to cortical thinning.

As a limitation, it has to be stated that non-heme-bound iron has a relevant impact on the paramagnetic tissue characteristics. Most human brain iron is found within the iron storage protein ferritin and its breakdown product, hemosiderin [61, 62]. For hypertensive patients, significantly higher iron content due to microbleeds has been reported [62]. In our study, patients with untreated hypertension were excluded both from the patient and the control group. Further, for patients and controls, there was no evidence of cerebral microbleeds in the acquired T2* maps. However, an additional impact of microbleeds on T2′ cannot be fully ruled out. Further, iron-rich amyloid plaques, which are contributors of non-heme iron in the aging brain [63], may have influenced T2′. However, the use of an age-matched control group should account for these age-related effects. Finally, it is worth noting that the more gradual reduction in oxygen consumption is accompanied by a slower increase in non-heme iron in the later decades [64]. T2′ should therefore be a useful tool to detect differences of deoxyHb in patients and controls of higher age.

Further, we lack additional supporting evidence from clinical correlations, since no neuropsychological testing was performed.

Conclusion

Non-invasive quantitative MRI with T2, T2*, and T2′ mapping may be used to detect structural and metabolic changes in ischemic leukoaraiosis. Assessing the grade of microstructural damage and hypoxia may be helpful to monitor disease progression and to perform a risk assessment.